A pulley is one of the easiest parts a printer can make. A round disc with a bore and a groove takes eleven minutes, needs no supports, and comes off the plate looking like the real thing. That is probably why so many printed drives run beautifully for a week and then quietly stop doing what they were designed to do.
The pulley is the easy part. The drive is not. A drive is a ratio you have to preserve, a belt length you have to live with, a tension you have to set once and hope holds, and two shafts that have to stay parallel under load. Get the disc right and miss those four, and the machine still fails. It just fails later, and more expensively.
This is the part of the machine that sits between the motor and the work. If you are still choosing the joint that grips the shaft, that is a separate problem, covered in hubs, shaft fits, keyways and couplings. If the shafts themselves are still undecided, see pins, pivots, shafts and bushings. Everything below is downstream of the shaft: how many teeth, how far apart, how tight, and what the plastic is actually doing under all of it.
First, the ratio is a decision, not a consequence
Most people arrive at a belt drive because they need to move something they cannot put a motor next to. They then pick a pulley that fits the available space and discover the output is either far too fast or far too slow. The ratio should be the first number you write down, before you open a CAD file.
For a belt drive, the ratio is just the tooth counts:
ratio = driven teeth / driving teeth
A 20-tooth pulley on the motor and a 60-tooth pulley on the output gives 3:1. The output turns one third as fast and, ignoring losses, carries three times the torque. That is the whole transaction. It is the same arithmetic as a gear train, and it fails in the same place: the plastic, not the math.
What the formula does not tell you is that the ratio has to hold while the belt is loaded, and that a drive with too much reduction in one stage becomes a drive with too few teeth in mesh. Reduction is cheap on paper and expensive in a pulley that strips.
What a printed drive actually has to survive
Before choosing teeth, know which failure you are designing against, because the six common ones pull in different directions. Only one of them is a strength problem.
| Failure signature | What you see | What it really is | Reversible? |
|---|---|---|---|
| Belt walk | Belt drifts to one flange and rubs | Shafts not parallel, or a crowned pulley pulling | Yes, realign |
| Tooth stripping | Teeth shear off one pulley | Too few teeth in mesh, or tension too low | No, rebuild |
| Bore slip | Pulley spins on the shaft; flats wear shiny | Friction grip where a shape was needed | No, rebuild |
| Flange rub | Grey dust, squeal, frayed belt edge | Belt too short, or an idler pushing sideways | Partly |
| Creep and ovalisation | A round pulley slowly becomes a cam | Constant side load on a plastic hub | No |
| Tension fade | Tension drops over weeks, then teeth skip | Elastomer belt relaxing, or the frame creeping | Re-tension once |
Notice that only tooth stripping is genuinely about material strength. Everything else is geometry, alignment, or the frame. That ordering matters, because the instinct is to reprint the pulley in the strongest filament available when the usual culprit is two shafts that are not parallel.
Tooth counts: the number that decides whether the belt survives
The most useful rule in belt drives is a minimum tooth count on the small pulley. Bend a belt around a very small pulley and the tooth roots on the inside of the curve take a compression the belt was never built for. Every belt standard specifies a minimum, and for the profiles used on printed drives it lands around 12 to 16 teeth on the driver. Below that, tooth-root fatigue starts and no amount of tension will save it.
| Driver teeth | Where it makes sense | What it costs you |
|---|---|---|
| 8 to 10 | Very low torque, short life only | Belt fatigue at the root, noisy, poor engagement |
| 12 to 14 | Light loads, intermittent motion | Workable, but watch tooth wear |
| 16 to 20 | The sensible default | Nothing. This is where drives live. |
| 24 to 40 | Reduction stages, output pulleys | Size and mass, but the belt is happy |
| Over 40 | Large reduction in a single step | Fewer teeth in mesh on the small side |
The companion rule is teeth in mesh. On the small pulley, count how many teeth the belt actually wraps. Below about six engaged teeth, the load per tooth climbs quickly and stripping becomes likely. A longer centre distance helps, and so does wrapping the belt more than halfway around the small pulley, which an idler can do deliberately.
Belt length: use the arithmetic, not the guess
Belt length is where printed drives get scrapped. You choose a centre distance that fits the frame, order a belt, and it is 30 mm too long. The formula for an open two-pulley drive is not complicated, and it belongs in your design notes:
L = 2C + 1.57 x (D + d) + (D - d)2 / (4C)
Here C is centre distance, D the large pitch diameter and d the small one, all in the same units. Pitch diameter is what the belt cares about, not outside diameter; for a toothed pulley it is roughly teeth x pitch / pi.
Worked example. A 20-tooth and a 60-tooth pulley on 2 mm pitch have pitch diameters of about 12.7 mm and 38.2 mm. At a 120 mm centre distance the belt comes out near 320 mm. You will not be able to order 320 mm, because belts come in fixed increments. So design it the other way round: pick a belt you can actually buy, solve for the centre distance, and build the frame's slots around that number with adjustment in both directions.
The part you should print first: the tensioner
Almost every printed belt drive needs a way to take up slack, and almost every printed belt drive is designed without one. If tension is set by the frame's exact length, then the frame's print tolerance, the belt's real length and the plastic's creep all add up in one place. The result is a belt that is either too tight on day one or loose by week three.
Three options, in the order I would trust them:
- Slotted motor mount with a jacking screw. Elongate the motor's mounting holes and add a screw that pushes the motor away from the driven pulley. Set tension by turning the screw, then lock the motor bolts. Build this first.
- Adjustable idler on a slot. It works, but an idler only tensions one span, and it adds a third shaft that also has to be parallel.
- Fixed centre distance, belt chosen to fit. Cheapest and fastest, fine for a prototype that will be rebuilt. Not a design for something that runs daily.
Whatever you pick, the tension itself needs a number rather than a feeling. For the small belts used on printed drives, a deflection test is enough: press the middle of the longest free span and measure how far it moves under a known force. A common target is roughly 1 to 2 percent of the span length under light thumb pressure. What matters more than the exact figure is that you use the same figure every time, so that "looser than last week" becomes something you can actually detect.
The idler, and why most of them do harm
An idler looks like a simple free-spinning pulley. In a printed drive it is usually doing one of three jobs, and only one of them was chosen deliberately.
| What the idler is for | How to get it right | The common mistake |
|---|---|---|
| Tensioning a loose belt | Bear on the slack side, push perpendicular to the span | Pushing on the tight side, which barely changes tension |
| Increasing wrap on the small pulley | Place it to wrap more than half the small pulley | Placed to wrap the large pulley, which was never the problem |
| Routing around an obstruction | Keep both bends on the same side where you can | An S-bend that reverses the belt and doubles the fatigue |
Two rules make idlers survivable. First, an idler must run on a bearing, not a printed hole riding a printed pin. An idler's surface speed is high, and a plain plastic bushing will heat, wear oval, and then start pushing the belt sideways. Second, an idler flanged on both sides is a belt-walk detector that has been disabled. Use one flange and let the belt find its own line, or use none and keep the run short.
If the idler rides a metal shaft, the interface is the same problem as any rotating joint, covered in printable bearings. If it rides a printed axle, the axle gets the same treatment as any other pivot, and the bending load arrives on the worst possible axis for a printed pin.
Pulley geometry: what actually has to be right
A printed pulley is three features stacked: a toothed rim, a web, and a hub with a bore. Each has one dimension that dominates.
| Feature | Dominant dimension | Why it dominates |
|---|---|---|
| Tooth rim | Tooth depth and root radius | Shallow teeth skip under load; sharp roots crack |
| Flange | Height above the belt, about 1.0 to 1.5 mm | A taller flange cannot stop walk, it only hides it until the belt is destroyed |
| Web | Thickness, roughly 1.6 to 2.4 mm | Too thin and the web dishes under tension, which changes effective pitch |
| Hub | Wall thickness around the bore | This is the torque path; thin walls ovalise and the pulley wobbles |
| Bore | Clearance to the shaft | A loose bore slips; a tight bore cracks the hub on assembly |
The flange point is counterintuitive enough to repeat. People add tall flanges to stop the belt walking. A tall flange stops nothing: the belt still walks, it just grinds against a taller wall and makes more dust before it fails. Flanges are the last line of defence for a drive that is already misaligned, not a fix for misalignment.
Orientation: print the pulley lying down, always
A pulley has one obviously correct orientation and one tempting wrong one. Lying flat on the plate, the bore is vertical, the layers form concentric rings around it, and the tension load from the belt is carried along the layer lines. Standing on its rim, the layers run across the bore and the belt load crosses them. That difference is worth roughly two to three times in the load the hub takes before it splits.
There is a second reason. A pulley printed flat has a bore made of stacked concentric circles, which is round to within the printer's circularity error. A pulley printed upright has a bore whose wall is a ladder of arcs and flats, because layer height becomes texture down the bore. Those flats are where the shaft wears first, and where a shaft will rock. The same logic applies to any hole in any part, and it is set out in inter-layer bonding and orientation.
Material: choose by tooth load, not by stiffness
The instinct is to reach for the stiffest filament in the drawer. For a pulley that is usually the wrong call, because a pulley's worst enemy is not deflection but tooth-root cracking under repeated load, and stiffness at the root is not the property that prevents it.
| Filament | Where it belongs on a drive | The caveat |
|---|---|---|
| PETG | The default for most printed drives | Tough at the root, but creeps under sustained side load |
| PETG-CF | Larger pulleys, longer centre distances | Stiffer, but the fibres make tooth roots more brittle |
| PA6-CF20 | The best all-round drive plastic | Hygroscopic; weak and stringy when wet |
| PA6-GF25 | Loaded drives where cost matters | Glass is gentler on tooth roots than carbon |
| PPS-GF20 | Anything near a hot motor | Needs a hot end that can reach it, and it is overkill elsewhere |
| ABS | Warm enclosures, cheap drives | Layer adhesion is the weak link at the root |
| PLA | Prototypes only | Creeps under constant belt tension, then ovalises |
| PLA Matte | Cover plates and guards on a drive | Cosmetic parts only; not a rim or a hub |
| TPU 95A | Not a pulley, but a belt face | Use it as a friction surface, never as a structural rim |
Two practical notes. Any fibre-filled filament needs a hardened or tungsten-carbide nozzle, such as the tungsten nozzle for Bambu Lab P1; a brass nozzle gets reamed out in a spool or two, and it happens gradually enough that you blame the model. Second, drying is not optional for nylon and is worth doing for PETG. A wet spool prints tooth roots with micro-voids, and the tooth root is exactly where the load arrives. See drying filament and the FilaBox drying box; a high-flow nozzle buys throughput, not dryness.
Getting the torque into the pulley
A pulley that slips on its shaft destroys the ratio silently. Everything downstream still turns, just not at the speed you designed, and the first symptom is usually a part that is wrong in a dimension nobody can explain.
Friction alone is a poor answer, for the same reason it is poor everywhere in printed mechanisms: grip depends on a bolt's preload, and the plastic under that bolt relaxes. If the drive does more than a light load, give the joint a shape. A D-flat, a keyway, a clamp collar or a metal insert all convert friction into geometry, and the trade-offs between them are laid out in hubs, shaft fits and couplings.
There is a drive-specific twist worth knowing. A pulley hub is loaded in torsion, and if the hub is printed in one piece with the rim, that torsion arrives at the web-to-hub junction. That junction, not the teeth, is where a printed pulley actually cracks. Give it a fillet and give it thickness, using the same reasoning as ribs, gussets and fillets.
Tight side and slack side are not symmetric
In a running drive the two spans carry different loads. The tight side carries the working tension plus the preload. The slack side carries the preload minus the working tension. The ratio between them is what decides whether the belt grips at all.
Three consequences follow. Idlers and tensioners belong on the slack side, where they change tension without fighting the working load. The tight side fails first, so keep that span short and unobstructed. And a drive with too little preload can have a slack side that drops near zero tension and starts to flap, which shows up as a rattle or a hum rather than as a failure.
That flapping gets misdiagnosed as a bearing problem more often than any other symptom. If your drive hums at one particular speed, check the slack span before you take anything apart; the diagnostics in rattles, hum and fatigue map onto a belt drive almost directly.
Alignment is most of the job
If you take one thing from this article, take this: a printed belt drive that walks its belt is almost never a pulley problem. It is two shafts that are not parallel, and no flange height fixes that.
Parallelism in two senses. The shafts must be parallel to each other in the belt's plane, and each must be perpendicular to the belt's direction of travel. A motor rotated two degrees about its mounting face will walk a belt off a pulley in a few hundred revolutions.
Checks, ordered by how much they catch:
- Lay a straight edge across both pulley faces and sight along it. Two degrees is visible by eye.
- Measure the centre distance at both ends of the shaft span. If the two readings differ, the shafts are not parallel.
- Turn the drive slowly by hand with the belt on and watch the belt's line against a fixed reference. The belt will tell you which way it wants to go.
- Only after alignment, ask whether the pulleys are crowned. A crowned pulley self-centres a flat belt and actively harms a toothed one.
The motor mount is part of alignment. A mount that flexes under belt tension tilts the motor and misaligns the drive dynamically, which is invisible to a straight edge at rest. Motor mounts, and the screw supports that keep them flat, are covered in printed motor mounts and rail brackets and in brackets that do not crack at the screw hole.
When a belt is the wrong answer
Belt drives are good at long centre distances and quiet running. They are bad at several other things, and a printed belt drive is worse at those than a metal one. A short reverse checklist:
- You need exact angular position. Belts stretch and have tooth clearance; use gears or a coupling.
- The centre distance is very short. Under about 40 mm there are too few teeth in mesh; use gears.
- You need high torque at low speed. Belt roots and pulley teeth both lose; add a gear stage.
- The environment is dusty or gritty. Belts carry abrasive into the grooves; see abrasion in sliding parts.
- You cannot tension it. Without adjustment, a belt drive is a bet on print tolerance.
- It runs continuously for months. Printed pulleys creep; a metal pulley on a printed bracket is usually the right hybrid.
The hybrid answer deserves to be stated plainly, because it is often correct engineering rather than a defeat. Metal pulleys, metal shafts, metal bearings, and a printed bracket and tensioner that hold them in place. The plastic does the geometry, which it is good at. The metal does the wear and the fatigue, which it is good at. It is the same split of duties as designing load paths rather than shapes.
Frame and bearing supports carry the side load
A belt drive puts a large side load on its shafts, and that load lands on whatever holds the bearings. In a printed frame this is the most commonly under-designed part of the whole assembly: a thin upright with a printed hole, carrying a load nobody sized it for.
Give the bearing a proper seat. A printed bearing pocket should be a press fit with a shoulder behind it so the bearing cannot walk axially, and the material around the pocket needs enough thickness that pressing the bearing in does not split the upright. If the bearing is a printed bushing instead, its length and clearance decide its life, and those rules are in printable bearings. If the drive sits near coolant or chips, sealing is a separate question, handled in sealing printed parts.
Stiffness matters more than strength here, because a support that deflects 0.2 mm under tension tilts the shaft and misaligns the drive. The section-efficiency logic in spars, ribs and hollow sections applies directly: a ribbed upright is far stiffer than a solid one at the same weight, and depth in the load direction beats extra thickness.
An assembly checklist you can actually run
Before the drive goes into service, these checks catch almost everything. They take an hour and need nothing beyond a caliper and a straight edge.
- Write down the ratio and the expected output speed, then measure it. If the measured speed differs, a slip is being hidden by friction.
- Check both pitch diameters with a caliper, not the outside diameter. A pulley printed slightly oversize shifts the effective ratio.
- Sight along a straight edge across both pulley faces. Fix alignment before anything else on this list.
- Set tension with the jacking screw, then run ten minutes and re-check. New belts settle, and frames creep on first load.
- Mark the belt and the pulley with a paint pen and confirm the marks move as the ratio predicts over twenty revolutions.
- Check for dust at the flanges after an hour. Any grey dust means the belt is rubbing, which means alignment or tension.
- Feel the bearing support with a finger after running. Warm is fine; hot means preload or misalignment.
- Re-check tension after a week, then after a month. If it drops twice, the frame is creeping and needs ribs, not more tension.
The rules, ordered by how much they matter
If the whole article had to fit on a card in the workshop, this is what would go on it.
| Priority | Rule | What it prevents |
|---|---|---|
| 1 | Shafts parallel before anything else | Belt walk, flange dust, most "pulley problems" |
| 2 | Tension by adjustment, never by frame length | Loose belts, skipped teeth, tension fade |
| 3 | Minimum 12 to 16 teeth on the driver | Belt root fatigue |
| 4 | Six or more teeth in mesh | Tooth stripping |
| 5 | Print the pulley flat, bore vertical | Split hubs, oval bores |
| 6 | Shape the shaft joint, do not rely on friction | Bore slip, silent ratio loss |
| 7 | Idlers on the slack side, on real bearings | Idler wear, sideways belt push |
| 8 | Belt length from the formula, centre distance from the belt | Belts that cannot be ordered |
| 9 | Ribbed, deep bearing supports | Deflection that misaligns the drive |
| 10 | PETG or PA6-CF by default, PLA only to prove the shape | Creep and ovalisation |
None of these are printing problems. Every one of them is a design decision that happens before the slicer opens, which is the same conclusion the rest of this series keeps arriving at: design rules beat slicer settings, the drawing decides the fit, and measuring is how you find out.
Once those ten are in place, a printed belt drive is a genuinely useful piece of machinery: quiet, cheap, forgiving of a slightly wrong centre distance, and easy to rebuild. It will not outlast a metal one. It will, with a jacking screw and a straight edge, do the job you actually needed it for.